This talk will also move through GPUs and other chips doing AI training in data centers- as many of them interconnect with each other. However, those interconnects do have only a limited bandwidth, which in turn limits AI training performance. Currently, a 2022 survey shows that the AI developers are allowed to use at most 25% capacity of the GPU.

Silicon photonics, says Xscape Photonics CEO and co-founder Vivek Raghunathan, could bring the answer in new interconnects with much-higher bandwidth. The secret sauce is that material: silicon-based, he explains, which manipulates light to transmit data.

"Xscape has created a platform that connects various computing elements in a sustainable way, yet offers the highest possible performance," Raghunathan said in an interview to TechCrunch. "The core of this platform's scaling relies on energy-efficient, cost-effective systems that do not exist in the industry yet," he said.

Headquartered in Santa Clara, at the heart of Silicon Valley, Xscape was born out of a lab at Columbia University, three professors: Alexander Gaeta, Keren Bergman, and Michal Lipson invented a technique they believed could be used to transmit terabytes of data over light.

In 2022, the three entrepreneurs spun off Xscape after bringing on board Raghunathan and Yoshitomo Okawachi, a laser engineer and a longtime colleague of Gaeta's. Raghunathan came aboard through Broadcom, where he cofounded the silicon photonics group, and Intel, where he was a manager over that company's silicon photonics products.

Conventional interconnects are metal wires that transmit data as electrical signals.

Metal-based interconnects consume a lot of power—and produce a lot of heat. They are bandwidth-limited by the conductivity of the media that they occupy. In addition, in those data centers when components are connected with fiber links, the interconnects' electrical data has to be converted into optical and back again, thus is latency-prone.

Silicon photonics, by contrast, use very little power and generate almost no heat.

In the past we mainly used optical communications for long-haul fiber-optic systems, said Raghunathan. But with recent advances, it's possible to integrate optics-on-chip in the form of silicon photonics and bring the optical interface all the way from the electronic plane into the chip.

Xscape's first product is a programmable laser to power data center fiber-optic interconnects, specifically the links among GPUs, AI chips, and memory hardware. The laser can leverage different colors of light (i.e. wavelengths) to transmit multiple data streams along the same link without interference, Raghunathan claims.

"Electrical systems densely packed together tend to produce crosstalk, interference and all these other challenges, but within the optical domain, the data can be modulated on different colors, wavelengths, or channels, and all co-propagate in the same wire or fibre, and don't interfere with each other."

This can be a major advantage over photonics competitors, such as Ayar Labs and Celestial AI. Xscape's lasers can be manufactured in the same facilities where microelectronics goes into phones and laptops. Assuming the tech works as advertised, Xscape faces the same challenge that most hardware startups face: manufacturing and selling its products at scale.

The first-generation laser can only emit between four and 16 colors, but Xscape already plans for improved versions of this technology that will be able to emit up to 128.

Xscape says it's "actively engaged" with 10 customers for potential deployments, ranging from vendors to hyperscalers — and that it has secured funding from Cisco and Nvidia, whose venture arms invested in its recent $44 million Series A round. The investments aren't strategic, meaning that the companies aren't currently customers. But Raghunathan notes that Cisco is one of the largest sellers of optical networking components in the world.

"This reflects Cisco and Nvidia's trust in the value we bring to this ecosystem," Raghunathan said.

The funding round was led by IAG Capital Partners, and brings the total the company has raised to date to $57 million. Funds will go toward growing the 24-person Xscape team, and scaling up fabrication of the lasers and related photonics tech, according to Raghunathan.

With this funding, Xscape will be able to extend its platform to integrate it with simulation, high-performance compute, and AI software and will enable customers in all industries to soar to new heights, Raghunathan said.

Xscape undoubtedly has a tough challenge ahead of it. Outside of Ayar and Celestial, the company competes with Intel in the multi-billion-dollar silicon photonics market. Intel claims it has shipped more than 8 billion photonics chips and 3.2 million on-chip lasers since 2016.